Stanford CME 295 - Transformers & Large Language Models
Here, you will find slides and recordings of class lectures, along with suggested readings.
Summary
A 3.5-hour general-audience lecture explaining how large language models like ChatGPT are built and work in practice.
It covers the full training pipeline, including data collection, tokenization, pre-training, fine-tuning, and reinforcement learning. The video also explains model behavior, limitations, and practical strategies for using LLMs effectively.